Effect of changing vegetation and precipitation on denudation – Part 1: Predicted vegetation composition and cover over the last 21 thousand years along the Coastal Cordillera of Chile

Vegetation is crucial for modulating rates of denudation and landscape evolution as it stabilizes and protects hillslopes and intercepts rainfall. Climate conditions and atmospheric CO 2 concentration ([CO 2 ]) influence the establishment and performance of plants and thus have a direct influence on vegetation cover. In addition, vegetation dynamics (competition for space, light, nutrients and water) and stochastic events (mortality and fires) determine the state of vegetation, response times to environmental perturbations, and the successional development. In spite of this, state-of-art reconstructions of past transient vegetation changes have not been accounted for in landscape evolution models. Here, a widely used dynamic vegetation model (LPJ-GUESS) was used to simulate vegetation composition/ cover and surface runoff in Chile for the Last Glacial Maximum (LGM), Mid Holocene (MH) and present day (PD). In addition, we conducted transient vegetation simulations from LGM to PD for four sites of the Coastal Cordillera of Chile at a spatial and temporal resolution adequate for coupling with landscape evolution models. Using a regionally-adapted parametrization, LPJ-GUESS was capable of reproducing present day potential natural vegetation along the strong climatic gradients of Chile and simulated vegetation cover was also in line with satellite-based observations. Simulated vegetation during the LGM differed markedly from PD conditions. Coastal cold temperate rainforests where displaced northward by about 5° and the tree line and vegetation zones were at lower elevations than at PD. Transient vegetation simulations indicate a marked shift in vegetation composition starting with the past-glacial warming that coincides with a rise in [CO 2 ]. Vegetation cover between the sites ranged from 13 % (LGM: 8 %) to 81 % (LGM: 73 %) for the northern Pan de Azucar and southern Nahuelbuta sites, respectively, but did not vary by more than 10 % over the 21,000 yr simulation. A sensitivity study suggests that [CO 2 ] is an important driver of vegetation changes and, thereby, potentially landscape evolution. Comparisons with other paleoclimate model driver highlight the importance of model input on simulated vegetation. In the near future, we will directly couple LPJ-GUESS to a landscape evolution model (see companion paper) to build a fully-coupled dynamic-vegetation/ landscape evolution model that is forced with paleoclimate data from atmospheric general circulation models.